Download Sponsler, Jeffrey, Frances Van Scoy, Doru Pacurari, 2002.

HPARSER
An Object Oriented Neuroanatomic
Atlas
Jeffrey L. Sponsler, MD MS, Frances Van Scoy, PhD, Doru
Pacurari, MS
West Virginia
University
Morgantown,
West Virginia,
United
States
Abstract
This report describes the design and development of a
knowledge based object-oriented system called the NeuroAnatomic
Atlas (NAA). The NAA
serves as the core of the Anatomic Localization Algorithm
(ALA) and encodes gross and macroscopic structures. Methods: A prototype atlas has
been implemented using the Common
Lisp Object System and consists of over 1100 objects including arteries,
cortical regions, basal ganglia,
brainstem nuclei, dermatomes, dorsal root
ganglia, and white matter tracts. The atlas has been associated with a
three-dimensional brain model for
display of structures. Results: Using a query language, the atlas provides a mechanism for
mapping from symptoms to anatomic pathways and from
pathways to vascular
structures. Fundamental localization algorithms have shown accuracy in finding
structural lesions. Keywords:
artificial
intelligence, object oriented design, neurology, expert systems,
computer graphics.
Introduction
This paper describes the design and development of the NeuroAnatomic Atlas (NAA). NAA employs object-oriented
knowledge-based techniques
to represent structures of the human central nervous
system.
Several neurology expert systems have been implemented using
artificial intelligence (AI) techniques. The SCAN system [1] and INKBLOT [2]
represented the nervous system using a hierarchy of cubes. Algorithms were
developed to reason about disease by analyzing geometric
relationships between
cubes. Smeets [3] developed a
system to prescribe anti-epileptic drugs with the conclusion “the system was at
least as good
in prescribing as individual neurologists.” A system called
NEUREX [4] diagnosed neurogenic diseases of the lower
limbs and helped to assist
users in planning electrophysiology tests.
We have addressed the knowledge representation of neuroanatomy from an object-oriented perspective [5] which we
consider an innovation. This
technology uses classes, methods, and inheritance
algorithms to organize data and has been applied to complex problems including
Hubble Space
Telescope scheduling ([6] and [7]). Our object-oriented model will
be comprehensive in its knowledge of macroscopic structures in the central
nervous system.
Project Goals
The goals of this development effort are listed here: (1)
Design and create the NeuroAnatomic Atlas (NAA) that,
once complete, may function as a
knowledge resource. (2) Write a set of
functions that encode the Anatomic Localization Algorithm (ALA). (3) Create a
three dimensional Brain
Model that illustrates the anatomic structures in the
atlas. (4) Test the atlas by analyzing correctness and potential for use in an
expert system
(described
briefly below).
The SYNAPS Expert System
Software called the System
for Neurologic Analysis of Patient Symptoms (SYNAPS) is in the early stages
of design and development. SYNAPS
will be a medical expert system focusing on
neurologic diagnoses and will accept, as input, patient signs, symptoms,
laboratory values, etc. A
hypothesis-driven rule system will produce a set of
diagnostic possibilities ranked by confidence.
SYNAPS will employ two key modules: NAA and the Anatomic Localization Algorithm (ALA).
ALA consists of a set of
functions that compute
the location of nervous tissue injury. These functions
query the atlas knowledge base and via heuristics derive localization.
Other potential applications for the atlas include teaching
systems, simulation, and model analysis.
Anatomy
Terms. Terms important to this report are defined briefly here. The neuron is a cell that functions as an
information processing unit. A
nucleus
is a collection of neuron cell bodies grouped by location and function. An axon is a fiber-like neuron component
that extends beyond the
neuron’s body to communicate with other neurons or with
muscle tissue. A synapse is a region
of chemical communication between two neurons.
A ganglion is a collection of neuron cell bodies usually found
outside of the central nervous system. The dermatome
is a region of skin innervated
by a spinal nerve. For more details, refer to [8],
[9], and [10].
Concepts for the Atlas
To design the atlas required definition of a set of concepts
that are described below.
Conceptual Threshold
Based on Size. A key decision was setting the smallest size for structures
to be included. Only structures above this size
threshold are defined. If an
inappropriate threshold is selected, the number of objects to be defined would
be too large for the initial phase of work.
Our threshold, at present, is on
the scale of millimeters and includes basal ganglia, brainstem nuclei, thalamic
nuclei, sensory tracts, motor tracts,
cortical regions (e.g., calcarine sulcus), Brodmann’s areas, dorsal root ganglia, etc. Microscopic
structures below the size threshold include
organelles, dendrites, individual
axons, neurotransmitter receptors, and synaptic vesicles.
file:///D/Libraries/Chad/My%20Documents/Dropbox/AKBC/ValleyNeuroscienceDreamweaver/papers/synaps-paper-2002.htm[12/7/2015 16:56:39]
HPARSER
CNS-Structures.
We begin model creation with the definition of cns-structure,
a discrete unit of excitable tissue possessing a specific biologic
function
(see [10]). A common cns-structure is the nucleus but others include cortical gyri,
sulci, areas, and lobes. Subclasses of the nucleus
include thalamic nuclei and the basal
ganglia. Each structure should be assigned arterial supply.
Brodmanns-Areas. This concept is a template for Brodmann’s areas that divide the cerebral cortex into discrete regions. Area
17 (the visual
cortex) is an example.
Connections.
The second key concept, the connection,
represents the axonal fibers between cns-structures.
Examples of cns-structures would be the
C5 dorsal root ganglion and the nucleus cuneatus.
The connection between these structures represents the set of axons with source
cell bodies in
the dorsal root ganglion and destination synapses in the nucleus
cuneatus. The class connection includes a source cns-structure,
a destination cnsstructure,
and one or more fiber tracts within
which the axons of the source are found. Generally, the cns-structure
is either a nucleus, a ganglion,
spinal cord gray matter region, or a cortical
area.
For example, a cortical
motor neuron whose body lies in the C5
motor area of the precentral gyrus
projects a long axon to the contralateral ventral
horn of the spinal cord. To represent this intercellular communication, three
objects are instantiated: (1) the motor area of the cortex
(corresponding to
the C5 spinal segment), (2) the ventral horn (at a specific spinal level C5),
and (3) the connection between them (with source
motor area C5 and destination
C5 ventral horn).
Subpathways. At the next
level of complexity, we define the concept of the subpathway. This class consists
of structures and connections of a
similar nature (e.g., sensory), associated
with a specific spinal cord level (e.g., 4th cervical level, C4).
The organization of the spinal cord dictates
the requirement for this concept
and the description is appropriate for sensory and motor tracts. Attributes
include modality (the physiologic
function), connections (a list of
connection objects), spinal cord level,
and laterality (left or right). The Left C4
Touch Subpathway is an example.
Processing of sensory signals in the dorsal horn, cuneate nucleus, thalamic ventral posterolateral
(VPL) nucleus, and sensory cortex is
physiologically complex [10]. We will, in
our initial phase, implement a system that maintains spinal cord sensory subpathways (e.g. C4) discretely
from dermatome to cortex.
This is a heuristic to support phased project completion. Later project phases
should address the integrative processing
known to occur in the sensory
pathways.
Pathways.
Composed of subpathways, the pathway concept defines a higher level nervous system complex. The
pathway is a set of cns-structures
and connections
that share a specific functionality such as the sensation of temperature.
Multiple spinal cord levels are included within one pathway
and a pathway is
unilateral. The definition of a pathway includes the major CNS structures that compose
the pathway.
For example, the Left Corticospinal System (which connects the left motor
cortex with the contralateral spinal cord ventral
horn) has a definition
that includes the left motor cortex, left internal
capsule, left crus cerebri,
corticospinal decussation,
right lateral corticospinal tract, and the right
ventral horn. Other pathways include the Right
Dorsal Column Medial Lemniscus Pathway (DCML) and
the Left Anterolateral
Pathway.
Systems.
The concept system encompasses a set
of two pathways (left and right). The semantic difference between a pathway and
a system is
merely that the system defines the biologic function for both sides
of the body. For example, the dorsal column medial lemniscus
(DCML) system
encompasses sensations of touch and vibration throughout the
entire body (not including that served by cranial nerves).
Arteries.
The class artery encodes components
of the vascular supply to nervous tissue. Knowledge about blood vessels is
crucial to analysis of
CNS disease processes. Our initial phase excludes veins and venous sinuses. The
attribute branches
includes pointers to the branches of the artery
instance (e.g., the
internal carotid artery is a branch of the common carotid artery). Ultimately,
every CNS structure will be assigned arterial supply.
System
Mapping Tables. Possessing definitions of a system such as the DCML
and the spinal cord level subpathway definitions, we
define
dermatome sensory mappings. To
adequately convert a symptom (e.g., paresthesia) to a
clinical diagnosis requires assignment of function to the
various systems,
structures, and connections already described. Our model accomplishes this by supporting
explicit definitions of an association
between an input and a subpathway. For example, an input would include the modality (e.g., touch), location (e.g., the C5 dermatome), and
the side
(e.g., left). The associated subpathway is
the C5-touch-subpathway-left. This subpathway is a component
of both the touch pathway and of the
DCML system.
Each of the concepts described above is associated with a
software class encoded using the Common Lisp Object System.
CNS
Object Table. The atlas requires a CNS Object Table that, for each entry, stores a unique key (i.e.,
the structure’s symbolic identifier) and an
object (that is an instance of a
class).
Knowledge
Base Definition Rules. NAA definition adheres to these rules: (a)
Definitions are coded in files and utilize functional specifications.
(b)
Definitions use unique symbolic identifiers for CNS structures. (c) Each
definition produces object instantiation and placement into the CNS
Object
Table. (d) Once definitions are complete, a function is executed that traverses
all objects and converts symbolic references to memory
pointers by table
lookup.
Nomenclature
and Laterality. The NAA uses the following naming convention. All symbols must be unique. Most identifiers
must include a
laterality suffix, “right” or “left.” Names usually follow this
example: The identifier “dermatome-C6-temperature.left” (representing the index
finger), is composed of “dermatome” (a skin region), “C6” (the spinal cord
level), “temperature” (the modality), and “left” (the side). When
subpathways
cross the midline, the side designation is chosen based on the side of the body
where sensation of motor activity occurs.
NeuroAnatomic Atlas Functions
Given the conceptual framework described above, a set of
functions has been written to embody these ideas. Two large groups, definition
functions
and query functions, are described below.
Definition Functions
Define-Nucleus
creates a nucleus object. Define-Connection
specifies the axonal connection between a source
structure and a destination structure
via a fiber path. Define-artery creates an artery object
with a list of branch identifiers. Define-cns-structure creates an object of class cns-structure.
file:///D/Libraries/Chad/My%20Documents/Dropbox/AKBC/ValleyNeuroscienceDreamweaver/papers/synaps-paper-2002.htm[12/7/2015 16:56:39]
HPARSER
Define-sensory-subpathway creates an object with an associated set of
connection identifiers.
Sensory-Subpathway: Light Touch
The following function defines the C4 touch sensory subpathway. The side effect of this function is the creation
of multiple objects in memory each
of which is cached by symbolic identifier
into the CNS Object Table. An example follows.
(define-sensory-subpathway
:id
'c4-touch.left
:component-of
'dorsal-column-medial lemniscus-system.left
:modality
'touch
:connections
'(c1 c2 c3 c4))
(define-connection :id 'c1
:source
'dermatome-c4-touch.left
:destination
'drg-c4-touch.left)
(define-connection :id 'c2
:source
'drg-c4-touch.left
:destination
'cuneate-nucleus-c4 touch.left
:tract
'dorsal-root-c4.left)
(define-connection :id 'c3
:source
'cuneate-nucleus-c4-touch.left
:destination
'vpl-c4-touch.right
:tract
'(internal-arcuate-fibers
medial-lemniscus.right))
(define-connection :id 'c4
:source
'vpl-c4-touch.right
:destination
'sensory-area-1-c4.right
:tract
'posterior-limb-internal capsule.right)
(define-dermatome
'dermatome-c4-touch.left)
(define-drg
'drg-c4-touch.left)
(define-nucleus 'cuneate-nucleus-c4 touch.left)
(define-tract ‘medial-lemniscus.right
:arteries
'(paramedian-rami.right))
(define-nucleus 'vpl-c4-touch.right)
(define-area
'sensory-area-1-c4.right)
(define-artery
'lateral-striate-artery.left)
(define-artery 'paramedian-rami.right)
Modality
Table. The modality table is a vehicle to associate specific nerve
modalities such as touch with a subpathway. The table
then provides a
mechanism to obtain an anatomic system from symptom
information. We define a function associate-symptom that populates the table. This
example creates the association of the sensory
modality of pain in the V1-face distribution (forehead) of
cranial nerve 5 (the trigeminal nerve) with
the subpathway v1-pain-pathway.
(associate-symptom
:sign
'pain :region 'v1-face
:modality
'face-v1-pain
:pathway 'v1-pain-pathway)
A second example defines a mapping from the C7 dermatome for
the sensory modality vibration to the specific pathway for this which is called
c7vibration.
(associate-symptom
:sign 'vibration :region
'c7
:modality
'vibration
:pathway
'c7-vibration)
Query Functions
To access the atlas, a set of functions to query the
knowledge base is defined.
Vessels-Of.
Let vessels-of be a function
definition that employs knowledge in the digital atlas to map from a structure
to its arterial supply. The
function returns a list of lists. Each inner list
contains the artery to the structure and recursively the parent arteries. So
for any given artery (e.g. the
lateral striate artery) we obtain the arterial
path back to the aorta. The algorithm follows:
LET Newlist = nil
LET Artery = artery-of (Cns-Structure)
LOOP
SET Parent = parent-artery
(Artery)
IF Parent == nil THEN exit
LOOP
ELSE push (Parent, Newlist)
SET Artery = Parent
END-LOOP
RETURN reverse (Newlist)
An example follows:
(vessels-of 'internal-capsule.left) =>
((<ARTERY lateral striate artery.left>
<ARTERY middle cerebral artery.left>
<ARTERY internal carotid artery.left>
<ARTERY common
carotid artery.left>
<ARTERY aorta>))
(vessels-of 'area-17.right) =>
((<ARTERY calcarine artery.right>
<ARTERY internal
occipital artery.right>
<ARTERY posterior
cerebral artery.right>
<ARTERY basilar
artery>
<ARTERY vertebral artery.right>
<ARTERY subclavian artery.right>
file:///D/Libraries/Chad/My%20Documents/Dropbox/AKBC/ValleyNeuroscienceDreamweaver/papers/synaps-paper-2002.htm[12/7/2015 16:56:39]
HPARSER
<ARTERY brachiocephalic
trunk.right>
<ARTERY aorta>))
Lookup-Pathway.
An operator Lookup_Pathway (modality body-region) is defined to
use the modality table. This operator takes as input a
symptom modality (e.g.,
pain) and a body region (e.g., V1-face) and produces an object that is the
pathway (e.g., V1-pain-pathway) stored. This is
the first step in the mapping
from symptom to anatomic structure.
Sign-To-Path-And-Arteries.
The function sign-to-path-and-arteries
provides important behavior for the automation of the diagnostic process by
querying the atlas, converting a symptom to a subpathway,
examining all connections, fiber tracts, and associated arterial supply. This
function has
the following behavior:
P = lookup-pathway (Modality Level Side)
LET ArteryList = nil
LOOP C over connections-of (P)
ArteryPath
= vessels-of (C)
PUSH (ArteryPath, ArteryList)
END-LOOP
RETURN ArteryList
Subpathway Traversal
Using the NAA and query functions, two examples of pathway
traversal and analysis are demonstrated.
Subpathway: C4-Touch
We may now trace the subpathway in
logical steps. Consider the example where the patient has loss of the sensation
of touch in the C4 dermatome
on the left side of the body.
(sign-to-path-and-arteries
'touch
'c4 :side 'left)
Via the modality table, we obtain the subpathway
for this dermatome and the list of connection objects for this object.
<SENSORY-SUBPATHWAY c4 touch.left>
CONNECTIONS: (C1 C2 C3
C4)
Each connection is analyzed to obtain artery information:
Process the first connection:
<Connection C1>
Source: <Dermatome C4 Left>
Destination: <GANGLION C4 Touch Left>
;; No artery information is available
Process the second connection:
<Connection C2>:
Source: <GANGLION-MODALITY C4 Touch Left>
Destination: <CUNEATE-NUCLEUS-C4 TOUCH.LEFT>
Tract: <DORSAL-ROOT dorsal root c4.left>
;; No artery information is available
Process the third connection:
<Connection C3>
Source: <NUCLEUS CUNEATE-NUCLEUS-C4 TOUCH.LEFT>
Destination: <NUCLEUS VPL-C4-TOUCH.RIGHT>
Tracts:
(<DECUSSATION
internal arcuate fibers>
<SENSORY-TRACT medial lemniscus.right>)
;; Fibers cross to contralateral side.
The function vessels-of reveals that there are two arteries
that supply the medial lemniscus:
((<ARTERY
<ARTERY
<ARTERY
<ARTERY
<ARTERY
<ARTERY
(<ARTERY
<ARTERY
paramedian rami.right>
basilar
artery>
vertebral artery.right>
subclavian artery.right>
brachiocephalic trunk.right>
aorta>)
anterior
spinal artery.right>
vertebral artery.right> …))
Process the fourth connection:
<Connection C4>
Source: <NUCLEUS VPL-C4-TOUCH.RIGHT>
Destination:
<AREA Sensory Area-1 C4.RIGHT>
Tract: <CAPSULE posterior limb internal
capsule.right>
Arteries of the internal capsule:
((<ARTERY lateral striate artery.right>
<ARTERY middle cerebral artery.right> …))
Subpathway: Pupillary
Constriction
The pupillary reflex subpathway
is described here. The retina can be divided reasonably into four quadrants. We
trace the connections of the left
upper quadrant of one retina to the mesencephalic nucleus controlling this reflex and the
motor path to the constricting muscles of the pupil.
<SENSORY-SUBPATHWAY pupillary
nasal upper quad1.right>
There are two connections:
(<Connection CS1>
<Connection CS2>)
Process the first sensory connection:
file:///D/Libraries/Chad/My%20Documents/Dropbox/AKBC/ValleyNeuroscienceDreamweaver/papers/synaps-paper-2002.htm[12/7/2015 16:56:39]
HPARSER
<Connection CS1>
SOURCE: <RETINA-NASAL-UPPER QUADRANT.RIGHT>
DESTINATION:
<NUCLEUS PRETECTAL NUCLEUS.LEFT>
TRACT: (OPTIC-NERVE.RIGHT OPTIC-CHIASM
OPTIC-TRACT.LEFT
BRACHIUM-OF-SUPERIOR-COLLICULUS.LEFT)
The arteries can be obtained via query.
Process the second sensory connection:
<Connection CS2>
Source: <NUCLEUS
PRETECTAL-NUCLEUS.LEFT>
Destination: <EDINGER-WESTPHAL NUCLEUS.LEFT>
Arteries of the Edinger Westphal Nucleus:
((<ARTERY paramedian rami.left>
<ARTERY basilar
artery> …))
The Edinger-Westphal nucleus
reacts to retinal light signals by producing pupillary
constriction. There are two connections in the motor segment of
this reflex:
<MOTOR-SUBPATHWAY
pupillary constriction.left>
CONNECTIONS:
(<Connection CM1>
< Connection CM2>)
Process the first motor connection:
<Connection M1>
Source:<NUCLEUS edinger westphal
nucleus.left>
Destination: <GANGLION ciliary ganglion.left>
TRACT: <OCULOMOTOR-NERVE.LEFT>
Process the second motor connection:
<Connection CM2>
Source: <GANGLION
CILIARY-GANGLION.LEFT>
Destination: <MUSCLE IRIS-SPHINCTER MUSCLE.LEFT>
This concludes traversal of the entire reflex path.
Algorithms for Localization
We propose a calculus
of disease localization that is based on the notion that central nervous system functional units may be geometrically
adjacent only in certain regions. If two nearby units are adversely affected, a disease process (tumor or stroke) may have occurred in the containing
structure. An example is the lateral medullary
syndrome wherein a stroke of the lateral medulla can be diagnosed by
findings that include
ipsilateral sympathetic
dysfunction and contralateral loss of pain and
temperature sense.
We have designed two techniques of localization: (1) the common arterial supply algorithm and
(2) the common parent structure algorithm.
They
are defined below.
Common Arterial Supply Algorithm
We define a function common-artery
that maps from symptoms to common arteries. Recall our function
vessels-of that processes a structure
argument and returns a list of
arteries. Let S1 and S2 be symptom variables.
Define-function Common-Artery (S1 S2)
intersection
(vessels-of
(symptoms-to-structure (S1),
vessels-of
(symptoms-to-structure (S2))
The resulting set of arteries should be ones that supply the
structures associated with the two different symptoms. The computed vessels
returned
may represent an area of vascular dysfunction. That is, if the common vessels
are disrupted (by injury or occlusion), their associated anatomic
structures
will be injured (by lack of oxygen), and the normal function of those
structures will be altered resulting in the symptoms experienced by
the
patient.
Common Parent Structure Algorithm
This algorithm requires a function adjacent-structure that returns true if its argument structures are
anatomically adjacent. The function componentof
(X, Y) returns true if structure X is an anatomic part of structure Y. This
example predicate component-of (nucleus cuneatus, medulla) is true. The
diagnostic heuristic common-parent-structure is defined here:
Let S1, S2, and PS be structures
Define-Function
Common-Parent-Structure (S1, S2, PS)
IF AND
(adjacent-structures (S1, S2),
component-of (S1, PS),
component-of (S2, PS))
THEN RETURN true
If dysfunction of S1 and S2 is found then the parent
structure PS may be injured as well.
Visualization of Structures
User interface design for the atlas includes brain
visualization in 3D and with virtual reality.
file:///D/Libraries/Chad/My%20Documents/Dropbox/AKBC/ValleyNeuroscienceDreamweaver/papers/synaps-paper-2002.htm[12/7/2015 16:56:39]
HPARSER
Scanned Human Brain Specimen
A human brain specimen (supplied by the West Virginia
University Department of Anatomy) was scanned using the Minolta VIVID 700
Scanner.
The resulting brain model was exported and manipulated using 3D
software. An illustration is included in this report.
Brain
Region Table. We have defined a table that supports a mapping from an
NAA object (“left face area of primary somatosensory
cortex”) to a
predefined area on the 3D Brain Model. An area is defined using
interactive tools to select polygons on the model. The set of polygons is
stored by
area name for later recall. A limited number of areas have been
defined and these include the somatosensory cortex
and the primary motor cortex.
Results and Discussion
The object-oriented NeuroAnatomic
Atlas has been partially implemented and includes over 1100 objects. As a
knowledge resource, this atlas
should support the SYNAPS expert system that is
currently in the design phase. The atlas functions as an
model and as a knowledge base of CNS
structures. The core knowledge is
sufficient to peruse the object network and to execute basic deductive analysis
of nervous system pathology.
Future Work
The System for Neurological Analysis of Patient Symptoms
(SYNAPS) is in the design phase now. The prototype NAA will provide (1) core
anatomic knowledge, (2) 3D visualization of structures, (3) a framework for
adding new paths and systems to the atlas, and (4) the core of teaching
software. We plan to augment the atlas with new structures, connections, and mappings
to areas on the 3D Brain Model. A natural language
software module [11] will be
used to convert free text information to expert system input data.
References
[1] Banks, Gordon, and Bruce Weimer, 1983. Symbolic coordinate anatomy for neurology (SCAN), in Proceedings of the Symposium on Computer
Applications in Medical Care 7 (1983), 828-830.
[2] Citro, G., G.
Banks, G. Cooper, 1997. INKBLOT: A neurological diagnostic decision support
system integrating causal and anatomical
knowledge, Artificial Intelligence in Medicine 10 (1997) 257-267.
[3] Smeets R, Talmon J, Meinardi H, Hasman A, 1999. Validating a
decision support system for anti-epileptic drug treatment. Part I: initiating
antiepileptic drug treatment Int J Med Inf 1999 Nov;55(3):189-98.
[4] Starita A, Majidi
D, Giordano A, Battaglia M, Cioni
R., 1995. NEUREX: a tutorial expert system for the diagnosis of neurogenic diseases of the
lower limbs, Artif Intell Med.
1995 Feb;7(1):25-36.
[5] Keene,
Sonya, 1989. Object-Oriented Programming
in Common Lisp: A Programmer’s Guide to CLOS,
Addison-Wesley, Reading.
[6]
Johnston, Mark, Glenn Miller, Jeff Sponsler, Shon
Vick, 1990. Spike: Artificial
intelligence scheduling for Hubble Space Telescope. Proc of
Fifth Conf on Artificial Intelligence for Space Applications (Huntsville,
May22, 1990), pp 11-18.
[7]
Sponsler, Jeffrey L., 1994. A Criterion Autoscheduler
for Long Range Planning, Proc. of the 1994 Goddard Conference on Space Applications
of
Artificial Intelligence (Greenbelt, Md).
[8] Adams,
Victor, Ropper, 1997. Principles of Neurology, 6th
Edition. McGraw-Hill, New
York.
[9]
Carpenter, Malcolm, 1991. Core
Text of Neuroanatomy 4th Edition,
Williams & Wilkins, Baltimore.
[10] Haines,
Duane, 1997. Fundamental Neuroscience,
Churchill Livingstone, New York.
[11] Sponsler, Jeffrey L. 2001. HPARSER: Extracting formal patient
data from free text history and physical reports using natural language
processing software. Proceedings of the American Medical Informatics
Association Annual Symposium, Washington DC,
Nov 3-11, 2001.
Acknowledgments
The authors wish to thank Dr. James Culberson (WVU Dept. of
Anatomy), Dr. John Brick (Chair, WVU Dept. of Neurology), and Brent Lally
(Dept. of Anatomy) for their support of this work.
file:///D/Libraries/Chad/My%20Documents/Dropbox/AKBC/ValleyNeuroscienceDreamweaver/papers/synaps-paper-2002.htm[12/7/2015 16:56:39]
HPARSER
Figure 1. Image
of the Neuroanatomic Atlas 3D Brain Model (from
scanned specimen).
file:///D/Libraries/Chad/My%20Documents/Dropbox/AKBC/ValleyNeuroscienceDreamweaver/papers/synaps-paper-2002.htm[12/7/2015 16:56:39]